The seeds of the future are in the present: A blind exploration of metastable states

TL;DR

This paper introduces a novel molecular dynamics simulation method that uses neural network-based biasing to discover new metastable states without prior knowledge, overcoming kinetic barriers in a blind manner.

Contribution

It presents a new blind sampling approach leveraging neural networks and a custom loss function to identify metastable states without prior reactive process knowledge.

Findings

Successfully discovers new metastable states in various examples.

Overcomes kinetic bottlenecks that hinder traditional methods.

Operates without prior knowledge of the system's reactive pathways.

Abstract

In this work, we present a novel type of molecular dynamics simulation that aims at discovering, in a blind way, new metastable states. Using only data coming from an initial unbiased simulation, and with the help of an appropriately defined loss function, we compute a bias that favors sampling yet unexplored configurational space regions, encouraging the system to leave the initial basin. In our work, we take advantage of what is normally thought to be a defect, namely the difficulty of neural networks to generalize. Contrary to most other enhanced sampling methods, which need previous knowledge of the reactive process, we are able to discover in a blind way new metastable states, overcoming otherwise insuperable kinetic bottlenecks. We illustrate the workings of the method with a number of instructive examples.